Database Reference
In-Depth Information
A
D
B
C
FIGURE 8.1
Links between URIs.
context, whereas D links to A for a different reason. (Note that these links need
not, and probably will not, be
owl:sameAs
, but other, more complex links like
“controls,” “is located in,” etc.). Therefore, although it is clearly better to specify
context explicitly, context can also emerge implicitly from links already present on
the Semantic Web.
At the macro scale, there are also links between entire datasets—the only thing
the modularization of data into a prepackaged dataset is really giving us is the
knowledge that all of the nodes within that dataset have the same context. It is worth
referring back to the vision of a properly joined Semantic Web, where there should
be little technical difference between an interdataset link and an intradataset link;
the only differences should be semantic.
8.8 LINK MAINTENANCE
Dealing with the situation of identifying and managing broken links is another rela-
tively unexplored area of Linked Data. Links tend to break when one or another
(and sometimes both) of the datasets change. Therefore, the most obvious approach
to avoiding the problem of broken links is via a system of notification when a data-
set changes. According to Leigh Dodds,
18
there are four categories of information
that could be notified: dataset notifications, when a new dataset has been added or
updated; resource notifications, which detail which resources have been added or
changed within the dataset; triple notifications, which provide information about the
individual triples that have changed; or graph notifications, when modifications have
been made to certain named graphs within the dataset. Notifications can be push or
pull. Pull mechanisms include subscribing to feeds, using Linked Data crawlers, or
querying datasets repeatedly to identify changes. Push mechanisms work by the data
consumer subscribing to a system to which the data provider publishes information
about its changes.
A number of ontologies have been published to describe the frequency of data
updates, for example, the Dataset Dynamics Vocabulary
19
or time-ordered sets of
update events such as the DSNotify Eventset Vocabulary (Popitsch and Haslhofer,
2010). While DSNotify takes a resource-centric view of changes, the Talis Changeset
